Re: Execute and Call_function of complex things

[Craig Markwardt <craigmnet(at)cow.physics.wisc.edu>]

dirk@uwast.astro.wisc.edu (Dirk Fabian) writes:

> ... Sorry that didn't post.
> 
> Ok, once again.  This goes in the "why doesn't it work like i think it 
> should" category.
> 
> Suppose I have an expression which, for the sake of argument, looks 
> something like: 
> 
> exp((-1.)*(0.0 + Gauss1(x,p(0:2)) + Gauss1(x,p(3:5))))
> 
> where gauss1 is another function (of x) and p is a set of input 
> parameters.
> 
> I want to evaluate this function, but can't seem to do it with either 
> call_function or execute.  I've tried:

Greetings to my alma mater and home state!

Pavel has it all right, so I just wanted to bring it all home.

You should use CALL_FUNCTION when you have a function *name*.  You
can't use an expression.

  y = call_function('sin', x)

You should use EXECUTE on a complete *statement*.  You can't pass any
arguments to an execute command aside from the command itself.  Thus
there really are no "local" variables.  Any variables that appear in
your statement must already be defined.

If you are using MPFITEXPR, which it looks like you may be doing, then
you have two choices.  Choice one is to enclose your expression in a
function named MODEL, and then use MPFITFUN instead, and evaluate your
model with this "Y = MODEL(X, P)".  The second choice is to use the
(documented but obscure) MPEVALEXPR which appears within the body of
MPFITEXPR. Like this, "Y = MPEVALEXPR(EXPR, X, P)"

Have fun!
Craig


-- 
--------------------------------------------------------------------------
Craig B. Markwardt, Ph.D.         EMAIL:    craigmnet@cow.physics.wisc.edu
Astrophysics, IDL, Finance, Derivatives | Remove "net" for better response
--------------------------------------------------------------------------